Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This paper studies the hierarchical hub location problem with single assignment for a three-level hub network and focus on several solution methods. The top level consists of a network of central hubs, directly connected with each other. The second and third level are star networks connecting each hub to exactly one central hub and connecting each node to exactly one hub or central hub. The objective is to decide on the location of the central hubs and the hubs, the allocation of hubs to central hubs and the allocation of nodes to hubs or central hubs, while minimizing the total distribution cost. Therefore, a linear mixed integer program formulation is used. Next, two models are proposed to select good values for the input parameters of the MIP formulation. The total solving time required for the full problem appears to be more than 14 hours for one set of number of central hubs and number of hubs to be opened. Therefore, this paper discusses three approximate heuristic solution methods, to find a nearly optimal solution in a much more acceptable time interval. Eventually, the performance of these heuristic solution methods is compared with the optimal solution of the full-sized problem.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

In deze masterthesis zijn twee data analyse technieken toegepast op parkeergegevens van de stad Alkmaar (NL) om patronen hierin te herkennen. De eerste techniek is datamining, waar een globaal beeld gegeven wordt over de data. Om deze techniek toe te passen is er gebruik gemaakt van de software Disco. Dit is een een open source tool waarmee patronen herkenbaar gemaakt kunnen worden. De tweede techniek die toegepast is, is Formal Concept Analysis (FCA) hiermee kunnen concretere patronen gevonden worden in een dataset. De software die hiervoor gebruikt wordt is Concept Explorer, ook deze tool is vrij te verkrijgen via het internet. In deze thesis worden enkel patronen gezocht in de gegevens van de bewoners omdat enkel deze van elkaar onderscheiden kunnen worden. Dit is de eerste keer dat er met deze analyse tools op parkeerdata gekeken wordt. Dit moet in het achterhoofd gehouden worden tijdens het lezen van deze tekst.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This thesis issues the planning and assignment of incoming flights and their inbound baggage to baggage conveyor belts in the baggage claim area of an airport. Inbound baggage handling represents a crucial process in the activities ongoing in the airport terminal. This directly affects the airport performance and the quality of service perceived by the passengers. Using a set of instances based on real data from Brussels Airport, a robust planning model is constructed that takes airline preferences, potential overlap and passenger satisfaction into account. It ensures minimal overlap whilst using existing infrastructure as efficiently as possible. First, a mathematical model of the problem is formulated. Its optimal solution serves as a benchmark for the implementation of two metaheuristics, namely Simulated Annealing and Tabu Search. The realistic data is featured in a comparative analysis between the general optimal solution of the mathematical model and both metaheuristics. The metaheuristics prove to be useful by delivering high-quality solutions in limited computational runtime. The results show that both Simulated Annealing and Tabu Search are capable of achieving a relative performance in the range of 95.66% up to 99.25%, compared to the optimal flight to baggage carousel assignment. The algorithms account for a planning horizon of three hours with a limited runtime of three minutes. Ultimately, metaheuristics can relieve dispatchers from daily planning operations in airports, making it more efficient.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The periodic vehicle routing problem is a well-studied problem in which routes for capacitated vehicles are planned over a planning horizon of several periods and customers have a set of allowed visit schedules. The periodic vehicle routing problem with driver consistency (PVRP-DC) is a new version of this problem in which each customer should always be visited by the same driver. In this thesis we illustrate the use and need of solutions to this problem in the industry through interviews with three different companies. We then go on to implement three simple meta-heuristics to solve the PVRP-DC: a Tabu-search, a Simulated Annealing search and a Genetic Algorithm. We report computational results on a wide range of instances and compare these to the integer linear programming solutions from [1].

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This work gives a broad understanding in how certain machine learning algorithms work and behave in the real world. The classification of the algorithm gives a structural representation of its functionality and risk for biases. The biases itself could be directly linked to the algorithm but also been implied by the data or even been introduced by loops in the system. These biases can lead ethical concerns, such as privacy. To accommodate this a complex legal structure is introduced, one of the most recent and important being the GDPR-law which is used in the European Union, on top of the local laws. To demonstrate the importance and real world application, the Schufa organisation is described. This credit scoring firm uses a black box which led to critique and ethical problems. Secondly the possibilities of recent algorithms, such as the support vector machine (SVM) is taken under the loop. This has led to a better credit assessment, and with future improvements, it should be possible to eliminate the current dilemmas. Lastly the wide applications of machine learning algorithm is noted by the example of Word2Vec. This is to indicate possible ethical dilemmas, such as gender biases. And more importantly how to eliminate them.